A type of integrated circuit widely used for micro electronic devices (e.g., processors and memories) is Complementary Metal Oxide Semiconductor (CMOS) which uses N-Channel MOS (N-MOS) and P-Channel MOS (P-MOS) devices or transistors built on the same substrate (FIG. 1). Such devices are often made on semiconductor substrates such as silicon wafers.
There are different crystal lattice orientations in a semiconductor substrate depending on the cut of the semiconductor substrate. Examples of several crystal lattice orientation include [001], [100], and [110]. Optimally, a CMOS device should be such that it has a high electron mobility for a high performance N-MOS device and a high hole mobility for a high performance P-MOS device. The mobility of electrons or holes depends significantly on the orientation of the crystal lattice of the semiconductor substrate. For example, for a device (e.g., a transistor) to have a high electron mobility, the channel of the transistor where electrons travel across should lie along a [001]-type plane. For a device (e.g., a transistor) to have a high hole mobility, the channel of the transistor should be parallel to a [110]-type plane. Thus, it is desirable to form N-MOS devices on [001] crystal planes to maximize the electron mobility and P-MOS devices on [110] crystal planes to maximize the hole mobility. Currently, as shown in FIG. 1, both P-MOS and the N-MOS devices are often made on the same semiconductor substrate (e.g., a 100-oriented silicon substrate) and thus the mobility for both the electrons and holes cannot be maximized. Under the current practice, manufacturers compensate for the low hole mobility in a substrate by making P-MOS devices bigger so that the drive current is relatively the same for both the N-MOS and the P-MOS devices made on the same substrate. As devices approach smaller and smaller dimension, compensating for the hole mobility by increasing the P-MOS dimension is impractical and undesirable.
Under the current practice, a dual orientation substrate (e.g., a substrate with a [001] orientation surface area and a [110] orientation surface area) is created by bonding two differently oriented silicon wafers together to form a silicon-on-insulator substrate using methods known in the art (e.g., using SMARTCUT, Bonded and Etch Back Silicon On Insulator (BESOI), or Separation by Implantation of Oxygen). FIG. 2 shows a [110] orientation silicon wafer being bonded to a [001] orientation silicon wafer with a silicon oxide (SiO2) film formed between the two wafers. Alternatively, a [001] orientation silicon wafer is bonded to a [110] orientation silicon wafer with a silicon oxide (SiO2) film formed between the two wafers (FIG. 3). Next, one wafer is then thinned (e.g., using Chemical Mechanical Polishing, CMP) as shown in FIG. 4 (certain area of the [110] orientation silicon wafer is thinned) and in FIG. 5 (certain area of the [001] orientation silicon wafer is thinned). Next, an epitaxial silicon film is formed on the wafer as shown in FIGS. 6-7. As shown in FIGS. 6-7, the substrate has an area of [001] orientated silicon and an area of [110] orientated silicon. The P-MOS device can then be formed on the [110] oriented silicon region and the N-MOS device can then be formed on the [001] oriented silicon region to form the device shown in FIG. 1.
The current practice generates material wastes and high cost in making a dual orientation substrate for the fabrication of N-MOS and the P-MOS devices on the same substrate. The processes of wafer bonding and the material wasted in these processes drive the cost of making the devices high. Additionally, the thickness uniformity of the device substrates is more difficult to control, for example, due to the accuracy limitation of the thinning process.